Table 13

Inconsistency due to changes in the Variational Autoencoder (VAE) for Stable Diffusion XL Turbo model and with EfficientNet v2small model.

Testing/TrainirιgStable Diffusion XL TurboAutoencoder with KL Loss (AKL) 1.5Autoencoder with KL Loss (AKL) 2Consistency DecoderVAE from Kandinsky 2.2VAE from Pixart-αVAE from SD 2.1VAE from SD TurboSD1.4 EMA fine-tuned VAESD1.4 MSE fine-tuned VAESDXL fine-tuned VAETiny AutoEncoder for SD
Stable Diffusion XL Turbo1.29%12.91%16.84%24.17%13.44%17.84%13.42%13.47%17.65%13.32%2.14%21.08%

or Create an Account

Close Modal
Close Modal